Machine learning and forecasting
Models for demand and load forecasting, recommendations, ranking, classification and anomaly detection. We train them on your data, build them into working systems and keep watch over their quality after launch.
Who it is for
Companies that have accumulated data on sales, customers, operations or equipment and make decisions by experience or in spreadsheets today: how much to buy, what to offer a customer, where to look for a problem.
What the client gets
A model built into your process or product through an API, with a description of the data, the quality metrics and the limits. We hand over the training and deployment code and the monitoring of the model’s quality in production.
Forecasting
Demand, load, sales and other indicators from historical data.
Recommendations and ranking
Choosing products, content and offers for each user.
Classification and anomaly detection
Sorting items into categories and spotting unusual transactions, events and readings.
Data preparation
Cleaning, labelling and data pipelines, without which a model will not run reliably.
Running models
Deployment, monitoring of quality and data drift, and retraining as the data changes.
Questions
How much data do we need?
It depends on the task. We start by analysing what you already have and say plainly whether it is enough for a useful model — or whether data has to be collected and labelled first.
Do you guarantee model accuracy?
No: a model’s quality depends on the data, so we do not guarantee metrics. We agree the metrics up front, measure them on your data, and iterations are part of the project.
What happens to the model after launch?
Data changes, and a model’s quality can decline. So we track metrics and drift and retrain the model when needed.